Investigating the Role of Artificial Intelligence in Psychotherapy

by Kristen Benito, PhD, psychologist, Rhode Island Hospital and Bradley Hospital

Since the pandemic, the demand for exposure therapy has outpaced the number of trained experts available to deliver it. Alarmingly, current statistics show that less than eight percent of patients in need have access to this essential treatment. An innovative research study at Bradley Hospital is pioneering the use of artificial intelligence (AI) to support patients in need of exposure therapy and clinicians who provide it.

Psychotherapy Success: The Need for Quality Monitoring

Known as The ACE Study, a modified acronym of its subject, this new study explores how AI can reduce the labor-intensive process of measuring quality in exposure therapy. 

Monitoring therapy quality is critical for evaluating psychotherapy clinical trials and ensuring the adoption of evidence-based interventions (EBIs) in mental health settings. However, existing quality measures are often costly and labor-intensive, limiting their practicality.

Without efficient and specific metrics, it is challenging to assess whether key quality components are being delivered effectively or to measure their impact. Natural Language Processing (NLP), a subset of AI, holds exceptional promise for automating quality coding.

Exposure therapy for anxiety offers an ideal testing ground for automated quality coding using NLP. This approach has both significant public health implications and a well-defined theoretical framework to guide quality measurement.

Study Design and Scope

Building on earlier pilot studies, the ACE Study will refine and test NLP-based automated quality coding using audio data from six existing clinical trials. These trials involve 1,286 patients and 12,050 sessions, spanning both research and community settings.

Collaborations and Stakeholder Engagement

The study leverages partnerships with leading institutions and stakeholders to enhance its efficiency and utility, including:

• National Institute of Mental Health (NIMH): Collaborating with the NIMH Machine Learning Core team, which specializes in anxiety research and deep learning methodologies, including NLP.
• Exposure Therapy Consortium (ETC): Partnering with this international community of researchers and clinicians to develop centralized tools for studying exposure therapy delivery and mechanisms.
• Diverse Stakeholders: Engaging patients, families, providers, agency leaders, payers, policymakers, and technology developers to:
  • Validate NLP-based coding against human coders.
  • Test predictive capabilities.
  • Integrate feedback to ensure findings are practical and applicable.

Early Results and Future Directions

Preliminary findings suggest that the automated quality coding process is highly effective. By reducing the labor-intensive nature of traditional quality coding, this AI-driven approach has the potential to scale rapidly and make evidence-based exposure therapy more effective and accessible. Future uses of this technology could include:

• Therapists receive feedback to help solve common problems and enhance session quality.
• Insurance providers gain cutting-edge evidence of treatment value.
• Families can access objective validation of therapy quality to help guide treatment decisions.

The ACE Study represents a groundbreaking effort to address the critical shortage of exposure therapy providers while advancing the use of AI to improve mental health care quality and outcomes.

Project partners are the University of Vermont and the National Institutes of Mental Health.